TextureCrop: Enhancing Synthetic Image Detection through Texture-based Cropping

2024-07-22

arXiv.org

AI Summary

Conventional preprocessing methods—such as resizing or center-cropping—discard high-frequency forensic artifacts in high-resolution synthetic images, degrading detection performance. To address this, we propose a lightweight, texture-aware adaptive cropping method that jointly leverages local frequency-domain analysis and texture saliency modeling to localize and preserve artifact-rich high-frequency texture regions, enabling non-uniform cropping. The method is fully plug-and-play: it requires no modification or retraining of pre-trained detectors and is compatible with both CNN- and Transformer-based architectures. Evaluated on Forensynths, Synthbuster, and TWIGMA, it improves AUC by 6.1% over center-cropping and 15% over resizing, with negligible memory overhead. Our core contribution is a texture-driven cropping paradigm that effectively mitigates artifact dilution in high-resolution images, substantially enhancing the robustness of existing synthetic image detection (SID) tools on large-scale inputs.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Natural Language Processing: Safety and Robustness

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web crawling and indexing
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

Generative AI technologies produce increasingly realistic imagery, which, despite its potential for creative applications, can also be misused to produce misleading and harmful content. This renders Synthetic Image Detection (SID) methods essential for identifying AI-generated content online. State-of-the-art SID methods typically resize or center-crop input images due to architectural or computational constraints, which hampers the detection of artifacts that appear in high-resolution images. To address this limitation, we propose TextureCrop, an image pre-processing component that can be plugged in any pre-trained SID model to improve its performance. By focusing on high-frequency image parts where generative artifacts are prevalent, TextureCrop enhances SID performance with manageable memory requirements. Experimental results demonstrate a consistent improvement in AUC across various detectors by 6.1% compared to center cropping and by 15% compared to resizing, across high-resolution images from the Forensynths, Synthbuster and TWIGMA datasets. Code available at https : //github.com/mever-team/texture-crop.

Problem

• Image Forgery Detection
• Large Image Processing
• Detection Accuracy

Innovation

• TextureCrop
• texture-focused cropping
• image forgery detection